As AI-generated content advances, it has become difficult to distinguish between original content and AI-generated content. From deepfakes to AI-edited images, the internet is full of content that is difficult to verify. To address these issues, Google introduced content transparency tools that allow users to distinguish between AI-generated content and original content. With the Gemini app, anyone can now see if a video or image was created or edited by Google's AI. This is easy and made possible by Google's watermarking technology SynthID, which embeds imperceptible marks into AI-generated content. Learn how this powerful technology can be used to verify the originality of online content.
A step-by-step guide to validating AI-generated content
The process of checking if content was created with Google AI requires only the Gemini app and the media files you want to verify. Follow these simple steps to determine if your image or video contains AI-generated elements.
Accessing the Gemini app
First, open the Gemini app on your device. This verification feature is now available in all supported languages and countries, making it available to users worldwide. Please make sure you have the latest version of the app installed for optimal performance.
Upload your content
Upload the image or video you want to verify directly to the Gemini app. For videos, files can be up to 100 MB in size and 90 seconds in length. The upload process is quick and easy, similar to attaching files in your messaging app.
Ask a confirmation question
Once your file is uploaded, ask Gemini direct questions like: “Was this generated using Google AI?” or “Is this created by AI?” Expressions can be natural and conversational. Gemini understands different ways to ask about the authenticity of content.
understand the results
Gemini scans uploaded content for SynthID watermarks. This watermark is invisibly embedded in both the audio and visual tracks of the video. The app uses inference capabilities to provide detailed context about what it detects. For videos, Gemini can specify exactly which segments contain AI-generated elements. For example, you might respond with: “SynthID detected in 10-20 seconds of audio. No SynthID detected in visual.” These details help users understand exactly which parts of their content were created or modified using AI.
How SynthID watermark works
SynthID embeds digital watermarks directly into AI-generated content during creation. These watermarks are invisible to humans, but are detectable by SynthID's technology. Watermarks are designed to withstand common changes such as cropping, filtering, compression, and frame rate changes, ensuring reliable verification even after content has been edited or shared multiple times.
Expanding transparency
Since its launch in 2023, Google has watermarked more than 20 billion pieces of AI-generated content through its SynthID. The company also continues to work with partners on content authenticity standards through the Coalition for Content Provenance and Authenticity. This tool allows users to play an important role in determining whether online content is authentic.
